import operator import Untils import matplotlib.pyplot as plt # BP神经网络 # 数据集: 列1:截距 1;列2:x坐标; 列3:y坐标 dataMat,classLabels = Untils.loadDataSet("student.txt") dataMat = mat(dataMat) classMat= mat(classLabels) # 数据归一化 dataMat = Untils.normalize(dataMat) # 绘制数据集坐标散点图 Untils.drawClassScatter(dataMat,classLabels,False) # m行数 n列数 m,n = shape(dataMat) labelMat = classMat.transpose() # 步长 alpha = 0.001 # 迭代次数 maxCycles = 500 #构成线性分割线 y=a*x+b: b:weights[0]; a:weights[1]/weights[2] weights = ones((n,1)) # 计算回归系数 weights for k in range(maxCycles): # 通过sigmoid函数返回结果,h是梯度计算的结果,是一个列向量 # h = logRegres2.sigmoid(dataMatrix*weights) h = 1.0/(1+exp(-dataMat*weights))
import operator import Untils import BackPropgation import matplotlib.pyplot as plt # 数据集 dataSet,classLabels = BackPropgation.loadDataSet("testSet2.txt") # 初始化时第1列为全1向量, studentTest.txt dataSet = BackPropgation.normalize(mat(dataSet)) # 绘制数据点 # 重构dataSet数据集 dataMat = mat(ones((shape(dataSet)[0],shape(dataSet)[1]))) dataMat[:,1] = mat(dataSet)[:,0] dataMat[:,2] = mat(dataSet)[:,1] # 绘制数据集散点图 Untils.drawClassScatter(dataMat,transpose(classLabels),False) # BP神经网络进行数据分类 errRec,WEX,wex = BackPropgation.bpNet(dataSet,classLabels) # 计算和绘制分类线 x,z = BackPropgation.BPClassfier(-3.0,3.0,WEX,wex) Untils.classfyContour(x,x,z) # 绘制误差曲线 X = linspace(0,2000,2000) Y = log2(errRec)+1.0e-6 Untils.TrendLine(X,Y)
DELTA = multiply(err,BackPropgation.dlogsig(HM,out)) wDelta = W.T*DELTA delta = multiply(wDelta,BackPropgation.dlogsig(hp,tau)) dWEX = DELTA*tauex.T dwex = delta*SampIn.T # 更新权值 if i == 0: WEX = WEX + eta * dWEX wex = wex + eta * dwex else : WEX = WEX + (1 - mc)*eta*dWEX + mc * dWEXOld wex = wex + (1 - mc)*eta*dwex + mc * dwexOld dWEXOld = dWEX dwexOld = dwex W = WEX[:,0:nHidden] # 重构dataSet数据集 dataMat = mat(ones((shape(dataSet)[0],shape(dataSet)[1]))) dataMat[:,1] = mat(dataSet)[:,0] dataMat[:,2] = mat(dataSet)[:,1] # 绘制数据点 Untils.drawClassScatter(dataMat,transpose(expected)) # 绘制分类线 # 绘制误差曲线 X = linspace(0,1000,1000) Untils.TrendLine(X,errRec)
DELTA = multiply(err,BackPropgation.dlogsig(HM,out)) wDelta = W.T*DELTA delta = multiply(wDelta,BackPropgation.dlogsig(hp,tau)) dWEX = DELTA*tauex.T dwex = delta*SampIn.T # 更新权值 if i == 0: WEX = WEX + eta * dWEX wex = wex + eta * dwex else : WEX = WEX + (1 - mc)*eta*dWEX + mc * dWEXOld wex = wex + (1 - mc)*eta*dwex + mc * dwexOld dWEXOld = dWEX dwexOld = dwex W = WEX[:,0:nHidden] # 重构dataSet数据集 dataMat = mat(ones((shape(dataSet)[0],shape(dataSet)[1]))) dataMat[:,1] = mat(dataSet)[:,0] dataMat[:,2] = mat(dataSet)[:,1] # 绘制数据点 Untils.drawClassScatter(dataMat,transpose(expected)) # 绘制分类线 # 绘制误差曲线 X = linspace(0,1000,1000) Untils.TrendLine(X,errRec)
import operator import Untils import BackPropgation import matplotlib.pyplot as plt # 数据集 dataSet,classLabels = BackPropgation.loadDataSet("testSet2.txt") # 初始化时第1列为全1向量, studentTest.txt dataSet = BackPropgation.normalize(mat(dataSet)) # 绘制数据点 # 重构dataSet数据集 dataMat = mat(ones((shape(dataSet)[0],shape(dataSet)[1]))) dataMat[:,1] = mat(dataSet)[:,0] dataMat[:,2] = mat(dataSet)[:,1] # 绘制数据集散点图 Untils.drawClassScatter(dataMat,transpose(classLabels),False) # BP神经网络进行数据分类 errRec,WEX,wex = BackPropgation.bpNet(dataSet,classLabels) # 计算和绘制分类线 x,z = BackPropgation.BPClassfier(-3.0,3.0,WEX,wex) Untils.classfyContour(x,x,z) # 绘制误差曲线 X = linspace(0,2000,2000) Y = log2(errRec)+1.0e-6 Untils.TrendLine(X,Y)
import operator import Untils import matplotlib.pyplot as plt # BP神经网络 # 数据集: 列1:截距 1;列2:x坐标; 列3:y坐标 dataMat,classLabels = Untils.loadDataSet("student.txt") dataMat = mat(dataMat) classMat= mat(classLabels) # 数据归一化 dataMat = Untils.normalize(dataMat) # 绘制数据集坐标散点图 Untils.drawClassScatter(dataMat,classLabels,False) # m行数 n列数 m,n = shape(dataMat) labelMat = classMat.transpose() # 步长 alpha = 0.001 # 迭代次数 maxCycles = 500 #构成线性分割线 y=a*x+b: b:weights[0]; a:weights[1]/weights[2] weights = ones((n,1)) # 计算回归系数 weights for k in range(maxCycles): # 通过sigmoid函数返回结果,h是梯度计算的结果,是一个列向量 # h = logRegres2.sigmoid(dataMatrix*weights) h = 1.0/(1+exp(-dataMat*weights))